Predicting Gas Quality with Deep Learning Techniques During Chemical Production

Author photo: Shin Kai


We've already started to see effective use of emerging technologies such as Industrial IoT (IIoT)-enabled remote condition monitoring predicting gas qualityand Big Data analytics for predictive maintenance and similar offline applications; but process engineers are interested in knowing if and how emerging technologies can be used to improve the actual production process and product quality, like predicting gas quality with deep learning techniques during chemical production.

In a recent pilot program, Mitsui Chemicals, Inc. and NTT Communications Corporation (NTT Com), the information and communications technology (ICT) solutions and international communications business within NTT Group, have successfully created a Deep Learning technique that accurately predicts the quality of gas products during production; 20 minutes before the final product is created.

As we learned in a recent press release from NTT Com, the technique is based on modeling the relationship between the different data sets sourced from raw materials feeding into the reactor; reactor conditions; and the trace gas impurities that represent gas product quality, expressed here as "X-gas."

The goal of this joint project between NTT Com and Mitsui Chemicals is to improve the accuracy of detecting abnormalities in product quality to improve operational efficiencies and product quality.  The two companies initiated the pilot project at one of Mitsui Chemicals’ gas production plants in 2015.

Using Deep Learning Algorithms for Predicting Gas Quality During Production

Chemical plants traditionally detect product quality issues by comparing production data with various benchmarks and by having experienced employees manually inspect the data.

In contrast, with this joint development by NTT Com and Mitsui Chemicals, Deep Learning algorithms analyze the data; automatically processing relevant factors to model and predict outcomes.

predicting gas quality

In the gas production process, these factors are represented by 51 types of real-time process data such as temperature, pressure, and flow. According to the two companies, the end result is a highly accurate forecast of the quality of the resulting gas products.  In the beta test at the Mitsui Chemicals plant, the two companies have succeeded in keeping discrepancies between the concentration of X gases predicted by this model and the actual concentration of X gases within +/-3 percentage points in full scale.

predicting gas quality

Detecting Process and Product Problems

By improving the prediction accuracy of X gases concentration by deploying this model, operators in chemical plants will also be able to detect faulty sensors or measuring instruments and accurately assess the current and likely future condition of the plant, as well as any anomalies in the chemical product. This should improve the accuracy of alerts, leading to safer and more stable operation and smarter plant maintenance.

Future Plans

Mitsui Chemicals is studying feasibility of applying this type of next-generation production technology to a variety of areas.  These include:

  • Smart plant maintenance
  • Making operations safer and more stable
  • Establishing optimum multi-grade production systems required in high value-added strategy, and
  • Sharing operational knowledge required in globalization.

The company plans to conduct further research into using next-generation production technology (including IIoT, Big Data, and AI) to enhance equipment reliability and operating efficiency, and continue to expand production technology infrastructures to enable chemical companies to respond flexibility to changes in the business environment.

In conjunction with Japan's Virtual Engineering Community, NTT Com has been conducting verification tests to develop cloud and network environments that deliver improved plant productivity and more efficient maintenance procedures.  While, to date, there are few cases in which IIoT data analyses deliver specific benefits during production, NTT Com and ARC Advisory Group both believe that this could be a key technology for improving production efficiency in the near future.

NTT Com plans to fine-tune the AI methodologies used to develop these latest techniques by using data sourced during production faults and from other plants.  This would allow it to expand the scope of the application and improve their overall accuracy. In the future, NTT Com will combine various elements of its IIoT and AI research and develop the resulting solutions under the company's own AI brand.

According to a company representative, NTT Com plans to perform additional research into ways to improve the operational efficiency of chemical plants through the use of AI models.  This includes helping prevent machine failures and identify quality abnormalities. The company also aims to leverage similar models in the development of its IIoT solutions.


Today's convergence of IT and OT makes this a very exciting time for everyone in the industrial community; technology suppliers, predicting gas qualityend users, and analysts alike.  This pilot project demonstrates a practical application of Deep Learning/AI in the chemical production process; one that in time is likely to be able to be applied in a real-time, online basis to a variety of chemical production processes.

ARC recommends that end users and suppliers alike closely follow these technology developments and -- as NTT Com and Mitsui Chemicals have done -- collaborate in research and pilot projects that further leverage advanced information and automation technologies to further improve process performance and safety.

To learn more about how your peers are leveraging predictive analytics and other advanced technologies to improve plant performance, please join us at the upcoming ARC Industry Forum in Orlando, Florida.


 If you would like to buy this report or obtain information about how to become a client, please Contact Us.

Keywords: NTT Communications, Mitsui Chemicals, Deep Learning, Artificial Intelligence, Trace Impurity Concentrations, ARC Advisory Group.

Engage with ARC Advisory Group

Representative End User Clients
Representative Automation Clients
Representative Software Clients